With the rise of Artificial Intelligence, Meta wants to go one step further than its competitors. Mark Zuckerberg’s company launched ImageBind, which is defined as an AI model that combines different senses, just like people do.
“Understands images, video, audio, depth, thermal and spatial motion”, indicates a Meta post. Six modalities in one.
If the rest of the technology only covers text, image, video and audio, ImageBind points further.
Introducing ImageBind by Meta AI: the first AI model capable of binding data from six modalities at once. This breakthrough brings machines one step closer to the human ability to bind together information from many different senses.
More on this new open source work ⬇️
—Meta AI (@MetaAI) May 9, 2023
Mark Zuckerberg’s commitment to Artificial Intelligence, leaving aside the Metaverse in recent months, is growing. ImageBind is the second project from the Meta Lab: the first was LLaMa (Large Language Model Meta AI), an open source AI tool for researchers.
LLaMa, following the style of OpenAI’s ChatGPT, allows training and improving natural language processing models.
More details on Meta’s new Artificial Intelligence tool, ImageBind
According to Zuckerberg’s company, ImageBind shares the entire sensory experience of humans, by learning a single embedding space that links multiple inputs.
“You can even upgrade existing AI models to support input from any of the six modalities, allowing audio-based search, cross-modal search, multi-modal arithmetic and cross-modal generation”. points out Goal.
But Meta does not want to set limits. The company claims that at some point they will be able to link more senses, such as touch, speech, smell, and fMRI signals from the brain.
“There is still a lot to discover about multimodal learning,” says Zuckerberg’s company. “The AI research community still has to effectively quantify scaling behaviors that appear only in larger models and understand their applications.”
ImageBind is available to developers in the open source Meta source, next.
